Repetition of complex frequency-modulated sweeps enhances neuromagnetic responses in the human auditory cortex

Hearing Research 282 (2011) 216e224

Contents lists available at ScienceDirect

Hearing Research journal homepage: www.elsevier.com/locate/heares

Research paper

Repetition of complex frequency-modulated sweeps enhances neuromagnetic responses in the human auditory cortex Christian F. Altmann a, b, c, *, Carsten Klein a, d, Linda V. Heinemann a, Michael Wibral e, Bernhard H. Gaese d, Jochen Kaiser a a

Institute of Medical Psychology, Goethe University, 60528 Frankfurt am Main, Germany Career-Path Promotion Unit for Young Life Scientists, Kyoto University, Yoshida Konoe-cho, Sakyo-ku, Kyoto 606-8501, Japan Human Brain Research Center, Graduate School of Medicine, Kyoto University, Kyoto 606-8507, Japan d Institute of Cell Biology and Neuroscience, Goethe University, 60528 Frankfurt am Main, Germany e Magnetoencephalography Unit, Brain Imaging Center, Goethe University, 60528 Frankfurt am Main, Germany b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 7 March 2011 Received in revised form 15 July 2011 Accepted 29 July 2011 Available online 5 August 2011

Frequency modulations (FM) play a decisive role in our everyday communication. To investigate the processing of FM direction we measured change-related auditory cortex responses with human magnetoencephalography. First, we tested for FM direction selectivity by presenting FM sweeps with the same FM directions in a repeated series (RS). These series were interrupted by a deviant with the opposite FM direction. Second, we tested for the representation of abstract rules and presented series of FM sweeps with alternating FM directions (AS). The AS series were interrupted by a deviant which was a repetition of the series’ last FM sweep but broke the alternating pattern. For the RS, the deviant did not evoke significant change-related responses in the auditory cortex. However, for the first stimulus after the deviant, significantly stronger responses compared to standards were observed bilaterally in the auditory cortex at about 200 ms after stimulus onset. For the AS, we observed a similar bilateral changerelated signal enhancement for a deviant FM sweep breaking the alternating series. Since this response enhancement occurred for both RS and AS even after a single FM sweep repetition, we conclude that these activities represent local signal enhancements rather than change-related responses due to abstract rule violation. In sum, our data indicate repetition enhancement due to spectro-temporal interactions between successive complex FM sweeps. These enhancement effects were observed for the first but not further repetitions suggesting a second-order repetition suppression of the initial repetition enhancement. Ó 2011 Elsevier B.V. All rights reserved.

1. Introduction Frequency changes convey important information in our everyday communicative life. For instance, processing of phonemes strongly depends on frequency modulation (FM; Liberman and Mattingly, 1989). It has been proposed that deficits in non-verbal auditory processing might underlie specific language impairment

Abbreviations: A, ascending frequency-modulated sweep; AS, alternating series; D, descending frequency-modulated sweep; Dev, deviant; EEG, electroencephalography; FM, frequency modulation; ISI, inter-stimulus-interval; MEG, magnetoencephalography; MMN, mismatch negativity; MMNm, magnetic counterpart of mismatch negativity; RS, repeated series; SD, standard deviation; Std, standard. * Corresponding author. Career-Path Promotion Unit for Young Life Scientists, Kyoto University, Yoshida Konoe-cho, Sakyo-ku, Kyoto 606-8501, Japan. Tel.: þ81 75 753 9295; fax: þ81 75 753 9281. E-mail address: [email protected] (C.F. Altmann). 0378-5955/$ e see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.heares.2011.07.008

or dyslexia (Tallal et al., 1997; Bailey and Snowling, 2002). In particular, deficient perception of slow frequency modulations has been implicated in reading impairments (Witton et al., 1998). For this reason, one therapeutic approach to improve dyslexic symptoms targets the processing of frequency changes by intensive training programs (Gaab et al., 2007). Given the importance of FM for our communication, the question arises whether specialized neural substrates are involved in the processing of these signals. Psychophysical studies have tested this by exposing subjects to specific FM directions and observed subsequent adaptation to this direction (Gardner and Wilson, 1979). These adaptation effects suggest the existence of featureselective channels processing FM direction. Further corroborating evidence has been obtained in neurophysiological studies of macaque auditory belt areas which contain neurons selective for the FM rate and the direction of FM tones (Tian and Rauschecker, 2004). Concurrently, a previous study has shown that a large

C.F. Altmann et al. / Hearing Research 282 (2011) 216e224

proportion of primary auditory cortex neurons in marmoset monkeys respond stronger to sinusoidal FM compared to pure tones (Liang et al., 2002). Moreover, these neurons showed selectivity for the stimulus modulation rate. To assess stimulus selectivity non-invasively at a mass neural scale, change-related paradigms can be employed. We previously conducted a two-tone adaptation experiment with human magnetoencephalography (MEG) to test for responses sensitive to animal vocalizations (Altmann et al., 2008). The observed amplitude reduction suggested a critical involvement of the P2m component, an evoked magnetic field deflection at about 200 ms after stimulus onset, in the representation of the spectral finestructure of complex sounds. However, in a recent MEG study, we observed repetition enhancement instead of adaptation when the same FM direction was presented consecutively (Heinemann et al., 2010). These findings were obtained employing a two-tone adaptation design with complex FM sweeps and short inter-stimulusintervals (ISI: 200 ms). Possibly, these repetition enhancement effects were caused by FM-specific interactions between the spectral and temporal properties of consecutively presented complex FM sweeps. In addition to the short-term interactions described above, interactions in sequences of tones at a longer time-scale have been reported. In particular, a repeating series of auditory stimuli interrupted by a deviant stimulus leads to a change-related response in the time range of 100 mse250 ms after stimulus onset, the socalled mismatch negativity (MMN) (Näätänen et al., 1978; Baldeweg, 2006). Using this approach, early MEG studies have described the magnetic counterpart of the MMN (MMNm) also for linear FM glides when the FM direction of the deviant differed from the standard (Sams and Näätänen, 1991; Pardo and Sams, 1993). While the MMN has proven itself as an adequate method to test sensory or memory representations in the cortex, it has also been shown that it can reflect higher-order abstract features of sound sequences. In particular, an earlier MEG experiment has shown an MMNm by presenting a series of descending Shepard tones and then either an ascending or a repeated tone as deviant (Tervaniemi et al., 1994). Shepard tones elicit the illusion of an endless pitch decrement with a series of complex tones containing several harmonic sinusoidal components that differ in frequency but share the overall spectral envelope that taper off at the low and high frequency ends (Shepard, 1964). In the study by Tervaniemi et al. (1994), both the ascending and the repeated tone elicited a mismatch response when presented after the series of descending Shepard tones suggesting that violations of higher-order rules are preattentively encoded. Similarly, in an EEG study where both local and global stimulus features were manipulated, MMN was mainly observed for global deviation (List et al., 2007). The extraction of abstract features goes beyond frequency-based rules as demonstrated by a further EEG study (Paavilainen et al., 2003). In this study, tone-pairs with ascending or descending frequency relations or with increasing and decreasing intensity relations were presented in series that were interrupted by an opposite frequency transition. MMN responses were elicited for both abstract frequency and intensity changes suggesting that higher-order rules are represented at the level of MMN generation. Evidence corroborating the notion that the MMN can reflect abstract rule formation has been provided by a previous EEG study (Carral et al., 2005). Here, we employ an MMN oddball paradigm to test for changerelated responses to FM stimuli in human auditory cortex that are: i) selective for FM direction, and ii) selective for higher-order attributes of the stimulus stream. First, to test the selectivity for FM direction in the auditory cortex, we presented series of repeated (RS: repeated series) FM sweeps with the same FM direction (e.g. ascending) and switched to series with a different FM direction (e.g.

217

descending). This followed the logic of a roving oddball paradigm, in which all employed stimuli can serve as both standard and deviant, thus investigating change processing regardless of the underlying physical stimulus features (Baldeweg, 2006; Haenschel et al., 2005; Altmann et al., 2007). In a classical oddball paradigm, series of standard tones are interrupted by infrequent deviant tones. In contrast, in the roving oddball paradigm a series of standard tones is interrupted by a new series of which the first tone serves as a deviant and subsequent repetitions as standards. In our stimulus series, we used logarithmic FM sweeps that were smoothed at the borders of their FM range to avoid sharp frequency transitions when FM sweeps with the same direction are repeated. Our second aim was to investigate sensitivity of change-related responses to higher-order attributes of the stimulus stream. To achieve this, we presented alternating series (AS) of descending and ascending FM sweep pairs and switched to a series in which the deviant had the same FM direction as the previous standard, but broke the series of alternately ascending and descending FM sweeps. If change-related responses mainly reflect local mismatch, we would expect increased magnetic responses for the deviant that follows an RS of ascending or descending FM sweeps (e.g. standards: ascending, deviant: descending). If on the other hand the higher-order global structure is reflected in change-related responses, we would expect increased magnetic responses for the deviant that breaks the AS, i.e. a sequence of up- and down FM sweeps (e.g. standards: ascending->descending, deviant: descending). 2. Materials and methods 2.1. Participants Thirteen healthy, right-handed (one ambidextrous) subjects participated in the experiment. All subjects had normal hearing abilities and had no history of otological, neurological or psychiatric disease as indicated by self-report. Two participants were excluded from further analysis due to excessive head motion and one subject due to excessive blink artifacts. Thus, data from ten subjects (six male) were analyzed. Average age was 25.1 years (range: 21e32 years). Each subject gave written informed consent to participate in the study. The experiment was performed in accordance with the ethical standards laid down in the declaration of Helsinki of 1964 and approved by the local ethics committee of the Goethe University Medical Faculty. 2.2. Apparatus and stimuli Stimuli were complex FM sweeps generated with a sampling frequency of 44.1 kHz and a duration of 100 ms. The complex FM sweeps consisted of four sinusoidal components separated by one octave in frequency. The sweeps started at 265, 530, 1061 and 2121 Hz and rose logarithmically with a modulation rate of 10 octaves per second to 530, 1061, 2121 and 4242 Hz. Sounds were either presented as ascending or descending in frequency. We applied a logarithmic Gaussian filter with a mean of 1050 Hz and a standard deviation of 0.59 octaves ([meanSD; mean þ SD] ¼ [698Hz; 1581Hz]) for a smooth limitation of the frequency range. Thus, these stimuli were a short, continuous, and glissando-like version of the Shepard illusion (Shepard, 1964). In its classical form, Shepard tones consist of several sinusoidal components spaced at octave intervals. Amplitudes are largest for the intermediate frequency range and small for the highest and lowest frequencies. The illusion of an endless pitch change is induced by shifting all sinusoidal components in frequency but keeping the

218

C.F. Altmann et al. / Hearing Research 282 (2011) 216e224

overall spectral envelope constant. For our experiment, these stimuli had the advantage that the frequency transition between consecutive sounds was relatively smooth. All stimuli were shaped by rising and falling 5 ms ramps and were presented binaurally via airconducting tubes with insert earphones (E-A-R-tone 3A, Aearo Corporation, Indianapolis, USA) at a comfortable sound intensity of approximately 80e85 dB(A). The air tubes had low-pass-filtering characteristics, and compared to a 1 kHz sound intensity was reduced by about 10 dB at 3 kHz, by about 20 dB at 4 kHz and 40 dB at 5.5 kHz.

serves as a deviant for the previous series, the last stimulus serves as a standard. The following two FM sweep series were presented intermixed (Figs. 1a and 2a): 1) a series of repeatedly (RS) ascending (AA..AA) or descending (DD..DD) sweeps, and 2) a series of alternately (AS) ascending and descending sweeps (AD..AD or DA.DA). For the RS, deviants were the first of a sequence of repeated FM sweeps with a direction different from the preceding sequence (Fig. 1a: ..AADD and ..DDAA). For the AS, deviants were the FM sweeps that repeated the last FM direction of an alternating FM series (Fig. 2a: ..ADDA and ..DAAD).

2.3. Procedure

2.4. MEG data Acquisition

The MEG experiment consisted of four to six experimental runs (five subjects completed six runs, four subjects five runs, and one subject four runs) with a duration of 6 min. During sound presentation, subjects were instructed to watch a silent movie. Each run consisted of 361 sound presentations and the inter-stimulus interval was 300 ms. In particular, two types of sound series were presented in which six to fourteen (mean: ten) stimuli were presented before a deviant occurred. In a roving oddball design, series of repeated stimuli are presented and the first stimulus in a series

The neuromagnetic signals were recorded using a whole-head MEG system (CTF-MEG, VSM MedTech Inc., Coquitlam, Canada) with 275 magnetic axial gradiometers with an average distance between the sensors of 2.2 cm. The signals were recorded at a sampling rate of 600 Hz with an anti-aliasing filter at 150 Hz. The final signals were obtained by computing synthetic 3rd order gradiometers in software (Acq version 5.4, VSM MedTech Inc., Coquitlam, Canada). The head position was determined at the beginning and the end of each recording with three head position

Fig. 1. Repeated series (RS). a) This diagram shows the stimulation during the RS. A series of repeatedly ascending (A, upper row) or descending FM sweeps (D, lower row) was interrupted by a series of descending (upper row) or ascending FM sweeps (lower row). Stdt0: last standard before the deviant (t0: serial position 0); Devt1: deviant sound; Devt2-5: FM sweeps following the deviant. b) Grand average waveforms for all MEG sensors. Frame color corresponds to the serial positions of the stimuli in a (Stdt0 e Dev5). c) Source dipole moments for the standard, the deviant and the sounds following the deviant sound. The upper row depicts the source waveforms for the left auditory cortex (AC), and the lower row those for the right AC. The dashed colored lines represent the difference waves between deviants and standards and the gray shaded area depicts the 99.9% confidence interval of the difference. The black dashed boxes mark the serial positions with significant differences according to our statistical criteria (i.e. the 99.9% confidence interval of the difference does not include 0 and the difference exceeds five baseline standard deviations for more than 10 ms). The waveform colors correspond to the serial positions in a). The vertical dashed gray lines indicate stimulus onset.

C.F. Altmann et al. / Hearing Research 282 (2011) 216e224

219

Fig. 2. Alternating series (AS). a) This diagram shows the stimulation during the AS. A series of alternately ascending (A) and descending (D) (upper row) or descending and ascending FM sweeps (lower row) was interrupted by an FM sweep that broke the series but started with a repetition of the last tone in the preceding series. Stdt0: last standard before the deviant (t0: serial position 0); Devt1: deviant sound; Devt2-5: FM sweeps following the deviant. b) Grand average waveforms for all MEG sensors. Frame color corresponds to the serial positions in a (Stdt0 e Dev5). c) Source dipole moments for the standard, the deviant and the sounds succeeding the deviant sound. The upper row depicts the source waveforms for the left auditory cortex (AC), and the lower row those for the right AC. The dashed colored lines represent the difference waves between deviants and standards and the gray shaded area depicts the 99.9% confidence interval of the difference. The black dashed boxes mark the serial positions with significant differences according to our statistical criteria (i.e. the 99.9% confidence interval of the difference does not include 0 and the difference exceeds five baseline standard deviations for more than 10 ms). The waveform colors correspond to the serial positions in a). The vertical dashed gray lines indicate stimulus onset.

indicator coils that were placed at the nasion and the preauricular points to ensure that head movements did not exceed 0.5 cm. Additionally, each individual subject’s head shape was recorded with a spatial digitizer (Zebris Medical GmbH, Tübingen, Germany). 2.5. MEG data analysis The continuous raw data were low-passed filtered with a zerophase filter at a cut-off frequency of 40 Hz. The evoked magnetic fields for each series type (RS, AS), serial position (standard, deviant and succeeding stimulus presentations : Stdt0, Devt1 to Devt5) and FM direction (ascending, descending) were averaged separately within an epoch from 100 ms before to 400 ms after sound onset. A prestimulus period of 100 ms before sound onset served as baseline. For further analysis, the evoked magnetic fields were pooled over the two FM directions (ascending, descending; RS: AA.AA þ DD.DD; AS: AD..AD þ DA.DA). Epochs which contained signal variations larger than 2.5 pT were excluded from averaging. After artifact rejection, on average about 92 trials (range 57e112) per series type (RS, AS) and serial position (standard, deviant and succeeding stimulus presentations: Stdt0, Devt1 to

Devt5) per subject remained for event-related averaging. The average number of trials analyzed was similar across serial positions (92e94 trials). Regional source estimation was conducted with BESA 5.2 software (MEGIS software, Gräfelfing, Germany). Source locations were computed for the evoked magnetic fields averaged across the last stimulus of a series, the deviant (the transition between series) and the following four stimulus presentations for individual subjects, separately for the RS and AS. Two non-symmetrical regional sources, located in the bilateral superior temporal lobes were used to model the evoked magnetic field. Locations of the two regional sources were fitted on the responses within the 20-m time window prior to the maximal global field power of the N1m response peaking at around 100 ms after sound onset (see for example Okamoto et al., 2010 for a similar approach). For the RS, the goodness-of-fit was on average 91% (SD: 7%) within the fitted period and for the AS 92% (SD: 4%). The source locations were approximatively transformed into the standard brain coordinate space described by Talairach and Tournoux (1988). This approximation was based on the individual subject’s head shape and reference points (nasion, left and right preauricular) and is thus less

220

C.F. Altmann et al. / Hearing Research 282 (2011) 216e224

accurate than a transformation to a standard space informed by individual brain anatomy. This approach was necessary as no individual brain anatomies were available. For MEG signals, a regional source consists of two equivalent current dipoles with the same location, but orthogonal orientations. The orientation of the first dipole of each regional source was set to match the direction of the maximum dipole moment of the averaged eventrelated magnetic fields. Since the second dipole of the regional source did not account for substantial activity, only the first dipole of the regional sources was analyzed. To compare dipole moments across serial positions (standard, deviant and successive stimulus presentations: Stdt0, Devt1 to Devt5), source time-courses were calculated for each individual subject, series type (RS or AS) and serial position (Stdt0, Devt1 to Devt5). For each subject and series type, the difference between the source waveforms of the deviants, the four stimuli following the deviant and the standard stimulus preceding the deviant were calculated. To test if these differences were statistically significant a non-parametric bootstrapping procedure (Efron and Tibshirani, 1994) based on 1000 iterations was applied to derive a 99.9% confidence interval of the mean difference between serial positions. Furthermore, for a conservative assessment of statistical significance, we applied the criterion that a meaningful difference should persist for at least 10 ms and that the mean difference should exceed the variability of the difference baseline ([-100; 0] ms before stimulus onset) by 5 standard deviations. Finally, to investigate differences in peak amplitude for specific evoked components (N1m, P2m), the N1m component of the source waveform was defined for each individual subject and each series type and serial position as the MEG peak response within a time range of 50e150 ms after stimulus onset (see for example Hoshiyama et al., 2007). Similarly, the P2m component was defined as the MEG source waveform peak response within 150e250 ms after stimulus onset (Hoshiyama et al., 2007). We statistically compared the N1m and P2m peak amplitudes by applying a repeated-measures ANOVA and post-hoc t-tests, corrected for multiple comparisons based on a sequential Bonferroniprocedure (Holm, 1979). 3. Results 3.1. Repeated series (RS) To evaluate change-related responses for FM direction changes, we tested differences between standard stimulation with repeated series of ascending or descending FM sweeps and deviants as shown in Fig. 1a. A sensor level global field power analysis is depicted in Supplemental Fig. 1a. The global field power across all sensors exhibited a significantly stronger MEG response to the deviant (Devt1) compared to standard (Stdt0) from 303 to 325 ms

after stimulus onset. Furthermore, the MEG response to the FM sweep following the deviant (Devt2) was significantly stronger compared to the standard (Stdt0) from 190 to 256 ms after stimulus onset. The evoked magnetic fields (Fig. 1b) were modeled by a pair of bilateral regional sources within the auditory cortex of each individual subject. The positions were on average at an approximated position of [x, y, z] ¼ [48, 10, 9] mm (SD:  [4, 11, 10] mm) in the right hemisphere and [x, y, z] ¼ [45, 12, 11] mm (SD:  [4, 11, 13] mm) in the left hemisphere according to the Talairach and Tournoux (1988) stereotactic coordinate system (see Fig. 3a). The dipole moments showed a first peak (N1m) averaged across serial positions at about 95 ms after stimulus onset in both hemispheres and a second inversely oriented peak (P2m) at about 231 ms after stimulus onset for the sound after the deviant (Devt2) in the right hemisphere and at about 216 ms in the left hemisphere. A similar but weaker second peak response was observed also for the deviant stimulus (Devt1). As shown in Fig. 1c, we observed a significantly (p < 0.001) stronger magnetic field response in the bilateral auditory cortex for the second FM sweep in the deviant series (Devt2) compared to standard (Stdt0). This component reached significance in the right auditory cortex between 126 and 311 ms and in the left auditory cortex between 150 and 278 ms after stimulus onset. For the fourth tone in the deviant series (Devt4), we additionally observed a significantly stronger magnetic response compared to standard (Stdt0) in the time windows 163e175 ms and 218e232 ms in the left hemisphere. In addition, we statistically compared the differences of the peak values of the N1m and P2m components between deviants and standards as shown in Fig. 4a. For the N1m component, a one-way repeated-measures ANOVA showed significant differences between serial positions for the right hemisphere (F[5,9] ¼ 4.21; p < 0.01) but not for the left hemisphere (F[5,9] ¼ 1.78; p ¼ 0.14). Post-hoc t-test comparisons did not show significant differences between deviants and standards. In contrast, for the P2m component we observed significant differences in both hemispheres (right hemisphere: F[5,9] ¼ 8.03; p < 0.001; left hemisphere: F [5,9] ¼ 6.76; p < 0.001). Post-hoc analysis showed significantly stronger P2m components for the second stimulus in the deviant series (Devt2) compared to standard (Stdt0) both in the right (t [9] ¼ 4.53; p < 0.01) and in the left hemisphere (t[9] ¼ 3.56; p < 0.05). To test whether the later P2m components were adequately modeled, we computed the dipole source locations for individual subjects for Devt2. The dipole positions were on average at an approximated position of [x, y, z] ¼ [45, 13, 9] mm (SD:  [7, 16, 8] mm) in the right hemisphere and [x, y, z] ¼ [47, 18, 9] mm (SD:  [9, 11, 13] mm) in the left hemisphere.

Fig. 3. Regional N1m dipole source locations for a single representative subject superimposed onto the subject’s anatomical MR image. LH: left hemisphere, RH: right hemisphere. The z-values report the z-coordinates of the slices for the left and right hemisphere separately according to the Talairach and Tournoux (1988) coordinate system.

C.F. Altmann et al. / Hearing Research 282 (2011) 216e224

a

RS right AC

left AC 30 Dipole moment [nAm]

Dipole moment [nAm]

30

N1m

20

10

P2m 0

b

Std

Dev

t0

t1

t2 t3 t4 Serial Position

10

AS left AC

P2m

Std

Dev

t0

t1

t2 t3 t4 Serial Position

t5

right AC 30 Dipole moment [nAm]

N1m

20

10

0

N1m

20

0

t5

30 Dipole moment [nAm]

221

P2m

Std

Dev

t0

t1

t2 t3 t4 Serial Position

t5

N1m

20

10

P2m 0

Std

Dev

t0

t1

t2 t3 t4 Serial Position

t5

Fig. 4. Average source moment peak amplitudes for N1m and P2m components obtained during the a) RS and b) AS. The peak amplitudes are plotted for different serial positions: t0 ¼ Stdt0: last standard before the deviant (t0: serial position 0); t1 ¼ Devt1: deviant sound; t2-t5 ¼ Devt2-5: FM sweeps following the deviant. Error bars represent the standard error of the mean.

Taken together, while the sensor level analysis showed some late change-related response for the repeated FM series, we did not observe a significant MMNm in the source level analysis. In contrast, an increase of the P2m amplitude in bilateral auditory cortex after a single repetition of the same FM direction (Devt2) was evident both at the sensor and source level. 3.2. Alternating series (AS) Our second aim was to test whether changes of abstract patterns, namely alternately ascending/descending FM series can elicit change-specific magnetic responses (Fig. 2a). Global field power analysis across all MEG sensors is depicted in Supplemental Fig. 1b. The global field power showed a significantly stronger MEG response to the deviant (Devt1) compared to standard (Stdt0) from 201 to 219 ms and 283 to 308 ms after stimulus onset. Similar to the RS, the evoked magnetic fields (Fig. 2b) were modeled by a pair of bilateral regional sources on a single subject level. The positions were on average at an approximated position of [x, y, z] ¼ [47, 8, 9] mm (SD:  [9, 10, 12] mm) in the right hemisphere and [x, y, z] ¼ [44, 11, 10] mm (SD:  [6, 11, 14] mm) in the left hemisphere (see Fig. 3b). The dipole moments showed a first peak (N1m) at about 95 ms after stimulus onset averaged across serial positions and a second inversely oriented peak (P2m) at about 220 ms after stimulus onset for the deviant (Devt1) for the right hemispheric source and at 228 ms for the left source. For the AS, we observed a significantly stronger MEG response for the deviant (Devt1) compared to standard stimulation (Stdt0) in the right auditory cortex between 175 and 265 ms and in the left auditory cortex between 161 and 296 ms after stimulus onset (Fig. 2c). We additionally found a significantly stronger response for

the second stimulus after the deviant (Devt2) compared to standard stimulation in the time range between 117 and 228 ms for the right and between 165 and 213 ms for the left auditory cortex. Again, we statistically tested the differences of the peak values of the N1m and P2m components as shown in Fig. 4b. For the N1m component, a one-way repeated-measures ANOVA showed significant differences between serial positions for the left hemisphere (F[5,9] ¼ 3.13; p < 0.05) but not for the right hemisphere (F[5,9] ¼ 2.00; p ¼ 0.10). Post-hoc t-tests revealed no significant differences between the deviants and the standard. The P2m component showed significant differences in both hemispheres (right hemisphere: F[5,9] ¼ 4.05; p < 0.01; left hemisphere: F[5,9] ¼ 10.64; p < 0.001). In the right hemisphere no significant differences between P2m peak amplitudes were revealed by posthoc analysis but in the left hemisphere larger P2m amplitudes for for the deviant (Devt1) compared to standard (Stdt0) (t[9] ¼ 4.39; p < 0.01) were observed. The P2m source positions for the deviant (Devt1) were on average at an approximated position of [x, y, z] ¼ [47, 11, 9] mm (SD:  [4, 12, 6] mm) in the right hemisphere and [x, y, z] ¼ [46, 16, 12] mm (SD:  [6, 11, 14] mm) in the left hemisphere. In summary, for the AS our analysis - both at the sensor and source level - revealed a significant increase of the P2m component, in particular in left auditory cortex, when an FM sweep was repeated and thus interrupted the alternating FM series (Devt1). 4. Discussion The present study investigated the processing of FM sweeps by measuring change-related MEG responses in the auditory cortex.

222

C.F. Altmann et al. / Hearing Research 282 (2011) 216e224

For a repeated series (RS) of ascending and descending FM sweeps, we did not observe a significant MMNm component in the auditory cortex comparable to an MMN/MMNm reported for pure tone series in numerous EEG and MEG studies (e.g. Näätänen et al., 1978, 1988; Sams et al., 1985; Tiitinen et al., 1993). For the stimulus following the first deviant, bilateral MEG responses in auditory cortex after about 126e150 ms were significantly stronger compared to the standard stimulus. When presenting a series of alternately ascending and descending FM sweeps (AS), a clear bilateral change-related MEG response started at about 161e175 ms for stimulation with a deviant that broke this series but was identical to the last standard tone. Taken together, these results suggest a local response enhancement after presentation of two consecutive FM sweeps with similar FM direction and frequency range, rather than a mismatch response by FM direction deviants alone or a more abstract pattern mismatch response. 4.1. MMNm for FM direction changes? It is remarkable that we did not observe a significant MMNm for changes of FM direction for the auditory cortical sources. The global field power averaged across all MEG sensors shown in the supplemental material indicates a rather late significant changerelated response from about 303 to 325 ms. That this response was seen at a rather late latency and for the global field power but not for the auditory cortical sources indicates a non-auditory origin. The MMN and its magnetic counterpart have been observed for numerous types of auditory change, such as frequency, duration or spatial location (for review see Näätänen et al., 2007). Robust MMN has also been described after changes of FM direction (Sams and Näätänen, 1991; Pardo and Sams, 1993). The lack e at least within the auditory cortex - of a significant MMNm in this study might be accounted for by the differences to previous studies, in particular the different stimulus structure and sound presentation timing. More specifically, in contrast to previous studies, we employed complex harmonic FM glides, and the inter-stimulus interval was rather short with 300 ms. Possibly, these differences in spectrotemporal stimulus structure and inter-stimulus interval may have led to MEG response interactions that counteracted the MMNm. Future studies that systematically vary the complexity of FM sweeps and the ISI between the stimuli within sound sequences may clarify the role of these parameters in generating a mismatch response. Further investigations could also increase the complexity of the stimulus set towards spoken language: for example, in tonal languages meaning is conveyed by modulation of the formant frequency (Thai: Kaan et al., 2007; Mandarin Chinese: Luo et al., 2007; Yang et al., 2008). Such tonal speech sounds share various characteristics with our stimulus set, namely frequency modulation and harmonic structure, thus similar enhancement effects may be observed for this ecologically relevant stimulus class. 4.2. MMNm for abstract rule violations? Besides testing the MMNm to repeated FM series, our second aim was to investigate the MMNm to abstract rule violations. Our finding that a repeated sound interrupting a series of sweeps that alternately change their FM direction elicits an enhanced MEG response could be taken as evidence for a higher-order representation of the alternating tone series. Evidence for such abstract rule representation has been described in previous EEG and MEG studies (Tervaniemi et al., 1994; Paavilainen et al., 2003). In addition, a previous EEG study has described the phenomenon of repetition negativity: a repetition within a constantly changing sound stream leads to an MMN-like response within 100e200 ms after stimulus onset (Horvath et al., 2001). Further investigation has

indicated that this response might indeed be an MMN to a repeated sound that interrupts a changing sound stream (Horvath and Winkler, 2004). This would be in accordance with our findings in the AS, but our data from the RS suggest that the repetition of a sound after an FM direction change leads to a response enhancement regardless of whether the preceding sounds were repeated or constantly changing. Thus, we suggest that for the complex FM sweeps and the ISI used in this study, the observed repetition enhancement was based on the interaction of an FM sweep and its predecessor. At a first glance, this repetition enhancement resembles the so-called repetition positivity. This is a component with increased positivity for repeated pure tones and has been demonstrated in a roving oddball EEG experiment with varying numbers of standard sounds (Haenschel et al., 2005). Repetition positivity was acquired from fronto-central electrodes from 50 to 250 ms after standard sound onset suggesting involvement of primary and non-primary cortical generators. However, in the current study the enhanced responses of the P2m component were seen for a single repetition of an FM direction but waned after subsequent repetitions. In contrast to repetition positivity, the repetition enhancement observed in this study did not increase with the number of repetitions rendering it unlikely that the two phenomena arise from similar neural processes. 4.3. Repetition enhancement Repetition enhancement has been observed in an early MEG study (Loveless et al., 1989) that also established that enhancement effects are particularly strong for short ISIs (w150 ms). Possibly, for short delays between sounds temporal integration across sound events is reflected by enhanced MEG responses (Loveless et al., 1996). In the primary auditory cortex of the awake macaque monkey, consecutive stimulation with pure tone-pairs has been shown to have facilitatory or inhibitory effects on neural responses to the second tone (Brosch and Scheich, 2008). The type of effect observed depended on the tones’ frequency attributes and the interval between them: tone-pairs that had a similar frequency and were in the range of the neuron’s best frequency led to inhibition, while tone-pairs with different frequency characteristics could result in facilitated responses, which were maximal when the frequency separation was about one octave and when the temporal separation amounted to about 100 ms. Similarly, Bartlett and Wang (2005) have shown both neuronal inhibition and facilitation for amplitude and frequency-modulated tones in awake marmoset monkeys that depended on the spectral and temporal structure of the tone sequences. In a previous study employing complex FM glides we reported enhancements after repeated stimulation with the same FM stimulus (Heinemann et al., 2010). These response enhancements occurred during a sustained MEG response from about 200 ms after stimulus onset with an orientation similar to the N1m component. Possibly, the enhancement effects reflected interactions between frequency separation and temporal distance between the consecutively presented sounds. These interactions were observed for FM sweeps but not for unmodulated complex tones, indicating that they are based on the spectro-temporal dynamics of the glissandolike FM sweeps. To bridge the gap between the response enhancement effects observed for pure tone sequences and complex FM sweeps, future studies could test MEG responses to tone sequences that consist of stimuli derived from a continuum between steady-state harmonic complex tones and harmonic complex FM sweeps. The previously observed MEG response enhancement after a single repetition of an FM sweep at about 200 ms after stimulus onset (Heinemann et al., 2010) is in line with the present findings. However, whereas the enhancements in the

C.F. Altmann et al. / Hearing Research 282 (2011) 216e224

223

previous study occurred for a sustained response with an orientation similar to the N1m, the response enhancements in the present study affected the P2m with opposite orientation. The main differences between the two studies were the longer inter-stimulus interval in the current study (300 versus 200 ms) and the stimulus sequences (six to fourteen repetitions versus two-tone series). Thus, the enhancement effects in the two studies occur within a similar time-frame, but it is unclear whether they are caused by similar processes but altered due to differences in the stimulation paradigm or whether they are indeed based on two different neural processes. Further research that systematically manipulates the ISI and stimulus sequence structure is needed to clarify this question.

4.6. Conclusion

4.4. Neural mechanisms of FM direction selectivity

We are most grateful for helpful comments on an earlier version of this manuscript from Torsten Baldeweg and for support by Deutsche Forschungsgemeinschaft (DFG AL 1074/2-1).

Facilitatory and inhibitory spectro-temporal interactions may be more than a mere epiphenomenon, but rather an essential mechanism for FM rate and direction selectivity. Neurophysiological studies on the auditory cortex of echo-locating bats have suggested that neural facilitation together with sideband inhibition and duration tuning might play an integral role in shaping FM sweep selectivity (Razak and Fuzessery, 2006, 2008). Inhibitory sidebands also play an important role for FM direction selectivity in rat primary auditory cortex (Zhang et al., 2003) where they are correlated to the topographic order of AI. Bats, rodents and primates have been suggested to share similar mechanisms to establish FM direction selective responses in the inferior colliculus (Pollak et al., 2011). Furthermore, bats and primates have both been suggested to employ combination-sensitivity as an integration mechanism to process echo-location calls but also vocalizations (Kanwal and Rauschecker, 2007). In particular, combinationsensitivity describes nonlinear enhancement of neural responses due to the spectral or temporal arrangement of stimuli and has been observed on the single cell level in primates (Rauschecker and Tian, 2000) and in bats (Kawasaki et al., 1988). 4.5. Second-order repetition effect We observed repetition enhancement after a single but not for further repetitions. Possibly, the repetition enhancement itself is subject to suppression effects due to extended repetition. A related type of second-order repetition effect has been observed for neural adaptation to auditory gratings in ferret primary auditory cortex (Shechter and Depireux, 2006). The authors suggested that the auditory cortex employs two coding strategies with different weighting of sensory context. Similarly, in the present study the process responsible for repetition enhancement possibly acts on a more local temporal scale but is affected by a second process that spans a wider context and leads to a decrement of enhancement after the first repetition. Interestingly, previous studies on cat primary auditory cortex have reported neurons that exhibit stimulus-specific adaptation on multiple time-scales (Ulanovsky et al., 2004). Stimulus-specific adaptation has been proposed to underlie the mismatch negativity observed in evoked potentials during oddball paradigms (Ulanovsky et al., 2003) and might play an important role in auditory scene analysis. More specifically, a study that investigated the responses of macaque auditory cortex neurons to ’B0 tones in an ’ABAB’ sound stream has shown response suppression in conditions when stream segregation typically appears (Fishman et al., 2001). The second-order repetition effect observed in the current study may be related to stimulus-specific adaptation with a long time constant and future studies could help to clarify if there is a relationship to stream segregation on a behavioral level.

In summary, we observed MEG response enhancements in the auditory cortex when FM sweeps were repeated after a change of FM direction regardless of whether they succeeded repeated or alternating FM sweep series. This effect was seen only for the first but not subsequent repetitions indicating a decrease of repetition enhancement with further repetitions. This suggests a local enhancement effect in which the responses to a complex FM sweep depend on the directly preceding FM sweep. Acknowledgments

Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.heares.2011.07.008. References Altmann, C.F., Bledowski, C., Wibral, M., Kaiser, J., 2007. Processing of location and pattern changes of natural sounds in the human auditory cortex. NeuroImage 35, 1192e1200. Altmann, C.F., Nakata, H., Noguchi, Y., Inui, K., Hoshiyama, M., Kaneoke, Y., Kakigi, R., 2008. Temporal dynamics of adaptation to natural sounds in the human auditory cortex. Cerebral Cortex 18, 1350e1360. Bailey, P.J., Snowling, M.J., 2002. Auditory processing and the development of language and literacy. British Medical Bulletin 63, 135e146. Baldeweg, T., 2006. Repetition effects to sounds: evidence for predictive coding in the auditory system. Trends in Cognitive Sciences 10, 93e94. Bartlett, E.L., Wang, X., 2005. Long-lasting modulation by stimulus context in primate auditory cortex. Journal of Neurophysiology 94, 83e104. Brosch, M., Scheich, H., 2008. Tone-sequence analysis in the auditory cortex of awake macaque monkeys. Experimental Brain Research 184, 349e361. Carral, V., Corral, M.J., Escera, C., 2005. Auditory event-related potentials as a function of change magnitude. Neuroreport 16, 301e305. Efron, B., Tibshirani, R.J., 1994. An introduction to the bootstrap. Chapman and Hall, New York. Fishman, Y.I., Reser, D.H., Arezzo, J.C., Steinschneider, M., 2001. Neural correlates of auditory stream segregation in primary auditory cortex of the awake monkey. Hearing Research 151, 161e187. Gaab, N., Gabrieli, J.D.E., Deutsch, G.K., Tallal, P., Temple, E., 2007. Neural correlates of rapid auditory processing are disrupted in children with developmental dyslexia and ameliorated with training: an fMRI study. Restorative Neurology and Neuroscience 25, 295e310. Gardner, R.B., Wilson, J.P., 1979. Evidence for direction-specific channels in the processing of frequency modulation. Journal of the Acoustical Society of America 66, 704e709. Haenschel, C., Vernon, D.J., Dwivedi, P., Gruzelier, J.H., Baldeweg, T., 2005. Eventrelated brain potential correlates of human auditory sensory memory-trace formation. Journal of Neuroscience 25, 10494e10501. Heinemann, L.V., Rahm, B., Kaiser, J., Gaese, B., Altmann, C.F., 2010. Repetition enhancement for frequency-modulated but not unmodulated sounds: a human MEG study. PLoS One 5, e15548. Holm, S., 1979. A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics 6, 65e70. Horvath, J., Czigler, I., Sussman, E., Winkler, I., 2001. Simultaneously active preattentive representations of local and global rules for sound sequences. Cognitive Brain Research 12, 131e144. Horvath, J., Winkler, I., 2004. How the human auditory system treats repetition amongst change. Neuroscience Letters 23, 157e161. Hoshiyama, M., Okamoto, H., Kakigi, R., 2007. Priority of repetitive adaptation to mismatch response following undiscriminable auditory stimulation: a magnetoencephalographic study. European Journal of Neuroscience 25, 854e862. Kaan, E., Wayland, R., Bao, M., Barkley, C.M., 2007. Effects of native language and training on lexical tone perception: an event-related potential study. Brain Research 1148, 113e122. Kanwal, J.S., Rauschecker, J.P., 2007. Auditory cortex of bats and primates: managing species-specific calls for social communication. Frontiers in Bioscience 12, 4621e4640. Kawasaki, M., Margoliash, D., Suga, 1988. Delay-tuned combination-sensitive neurons in the auditory cortex of the vocalizing mustached bat. Journal of Neurophysiology 59, 623e635.

224

C.F. Altmann et al. / Hearing Research 282 (2011) 216e224

Liang, L., Lu, T., Wang, X., 2002. Neural representations of sinusoidal amplitude and frequency modulations in the primary auditory cortex of awake primates. Journal of Neurophysiology 87, 2237e2261. Liberman, A.M., Mattingly, I.G., 1989. A specialization for speech perception. Science 243, 489e494. List, A., Justus, T., Robertson, L.C., Bentin, B., 2007. A mismatch negativity study of local-global auditory processing. Brain Research 1153, 122e133. Loveless, N., Hari, R., Hamalainen, M., Tiihonen, J., 1989. Evoked responses of human auditory cortex may be enhanced by preceding stimuli. Electroencephalography & Clinical Neurophysiology 74, 217e227. Loveless, N., Levanen, S., Jousmaki, V., Sams, M., Hari, R., 1996. Temporal integration in auditory sensory memory: neuromagnetic evidence. Electroencephalography & Clinical Neurophysiology 100, 220e228. Luo, H., Boemio, A., Gordon, M., Poeppel, D., 2007. The perception of FM sweeps by Chinese and English listeners. Hearing Research 224, 75e83. Näätänen, R., Gaillard, A.W., Mäntysalo, S., 1978. Early selective-attention effect on evoked potential reinterpreted. Acta Psychologica 42, 313e329. Näätänen, R., Sams, M., Alho, K., Paavilainen, P., Reinikainen, K., Sokolov, E.N., 1988. Frequency and location specificity of the human vertex N1 wave. Electroencephalography and Clinical Neurophysiology 69, 523e531. Näätänen, R., Paavilainen, P., Rinne, T., Alho, K., 2007. The mismatch negativity (MMN) in basic research of central auditory processing: a review. Clinical Neurophysiology 118, 2544e2590. Okamoto, H., Stracke, H., Lagemann, L., Pantev, C., 2010. Bottom-up driven involuntary auditory evoked field change: constant sound sequencing amplifies but does not sharpen neural activity. Journal of Neurophysiology 103, 244e249. Paavilainen, P., Degerman, A., Takegata, R., Winkler, I., 2003. Spectral and temporal stimulus characteristics in the processing of abstract auditory features. Neuroreport 14, 715e718. Pardo, P.J., Sams, M., 1993. Human auditory cortex responses to rising versus falling glides. Neuroscience Letters 159, 43e45. Pollak, G.D., Gittelman, J.X., Li, N., Xie, R., 2011. Inhibitory projections from the ventral nucleus of the lateral lemniscus and superior paraolivary nucleus create directional selectivity of frequency modulations in the inferior colliculus: a comparison of bats with other mammals. Hearing Research 273, 134e144. Razak, K.A., Fuzessery, Z.M., 2006. Neural mechanisms underlying selectivity for the rate and direction of frequency-modulated sweeps in the auditory cortex of the pallid bat. Journal of Neurophysiology 96, 1303e1319. Razak, K.A., Fuzessery, Z.M., 2008. Facilitatory mechanisms underlying selectivity for the direction and rate of frequency modulated sweeps in the auditory cortex. Journal of Neuroscience 28, 9806e9816.

Rauschecker, J.P., Tian, B., 2000. Mechanisms and streams for processing of "what" and "where" in the auditory cortex. Proceedings of the National Academy of Sciences 97, 11800e11806. Sams, M., Näätänen, R., 1991. Neuromagnetic responses of the human auditory cortex to short frequency glides. Neuroscience Letters 121, 43e46. Sams, M., Paavilainen, P., Alho, K., Näätänen, R., 1985. Auditory frequency discrimination and event-related potentials. Electroencephalography and Clinical Neurophysiology 62, 437e448. Shechter, B., Depireux, D.A., 2006. Response adaptation to broadband sounds in primary auditory cortex of the awake ferret. Hearing Research 221, 91e103. Shepard, R.N., 1964. Circularity in judgments of relative pitch. Journal of the Acoustical Society of America 36, 2346e2353. Talairach, J., Tournoux, P., 1988. Co-Planar Stereotaxic Atlas of the Human Brain: 3Dimensional Proportional System: An Approach to Cerebral Imaging. Thieme, New York. Tallal, P., Miller, S.L., Jenkins, W.M., Merzenich, M.M., 1997. The role of temporal processing in developmental language-based learning disorders: research and clinical implications. In: Blachman, B.A. (Ed.), Foundations of Reading Acquisition and Dyslexia: Implications for Early Intervention. Lawrence Erlbaum, Mahwah, NJ, pp. 49e66. Tervaniemi, M., Maury, S., Näätänen, R., 1994. Neural representations of abstract stimulus features in the human brain as reflected by the mismatch negativity. Neuroreport 5, 844e846. Tian, B., Rauschecker, J.P., 2004. Processing of frequency-modulated sounds in the lateral auditory belt cortex of the rhesus monkey. Journal of Neurophysiology 92, 2993e3013. Tiitinen, H., Alho, K., Huotilainen, M., Ilmoniemi, R.J., Simola, J., Näätänen, R., 1993. Tonotopic auditory cortex and the magnetoencephalographic (MEG) equivalent of the mismatch negativity. Psychophysiology 30, 537e540. Ulanovsky, N., Las, L., Nelken, I., 2003. Processing of low-probability sounds by cortical neurons. Nature Neuroscience 6, 391e398. Ulanovsky, N., Las, L., Farkas, D., Nelken, I., 2004. Multiple time scales of adaptation in auditory cortex neurons. Journal of Neuroscience 24, 10440e10453. Witton, C., Talcott, J.B., Hansen, P.C., Richardson, A.J., Griffiths, T.D., Rees, A., Stein, J.F., Green, G.G., 1998. Sensitivity to dynamic auditory and visual stimuli predicts nonword reading ability in both dyslexic and normal readers. Current Biology 8, 791e797. Yang, L.J., Cao, K.L., Wei, C.G., Liu, Y.Z., 2008. Auditory pre-attentive processing of chinese tones. Chinese Medical Journal 121, 2429e2433. Zhang, L.I., Tan, A.Y.Y., Schreiner, C.E., Merzenich, M.M., 2003. Topography and synaptic shaping of direction selectivity in primary auditory cortex. Nature 424, 201e205.